Título: An Oscillatory Correlation Model for Semi-Supervised Classification
Autores: Quiles, Marcos G.; Basgalupp, Márcio P.; Barros, Rodrigo C.
Resumo: This paper presents a new semi-supervised classification algorithm based on the oscillatory correlation theory. In this approach, the dataset is converted into a network whose nodes represent the samples and the edges represent the similarity among these samples. Each node in the network is modeled by an oscillator. The network clustering is given by the oscillators synchronization phenomenon, whereas the separation of oscillators that represent distinct clusters is induced by a global inhibitor. The previously labeled objects make use of the synchronization dynamics in order to propagate labels among their neighbors. Experiments performed with the proposed approach have shown promising results in a variety of datasets. It has shown to be capable of eventually outperforming traditional methods in the literature.
Palavras-chave: Oscillatory correlation; synchronization; semi-supervised learning
Páginas: 8
Código DOI: 10.21528/lmln-vol11-no1-art1
Artigo em PDF: vol11-no1-art1.pdf
Arquivo BibTex: vol11-no1-art1.bib